Updated WFC3/IR Photometric Calibration...2020/12/02 · Calibration V. Bajaj, A. Calamida, J. Mack December 2, 2020 Abstract We present the continued analysis of photometric measurements

WFC3 Instrument Science Report 2020-10

Updated WFC3/IR PhotometricCalibration

V. Bajaj, A. Calamida, J. Mack

February 15, 2021

Abstract

We present the continued analysis of photometric measurements of the CALSPEC standardstars over the last 11 years in all of the 15 WFC3/IR filters. In general, the photometry(countrate) is consistent with the 2012 calibration to 1% or better. However, new modelsfor the CALSPEC primary white dwarfs changed the HST photometric flux referencesystem, thus changing the inverse sensitivities and zeropoints. This change is less than0.5% on average for the wide filters, but increases to just under 2% for the reddest mediumand narrow filters. No discernible changes in sensitivity over time are detected in themeasurements, but this is partially due to a lack of precision likely caused by persistence ofprevious observations, as well as other effects that are not currently well understood. Thenew zeropoint tables are presented here.

Introduction

The monitoring of the Wide Field Camera 3 Infrared (WFC3/IR) channel’s photometricperformance via the measurement of standard spectrophotometric stars has been continuedsince the launch of the instrument in 2009. The first 1.5 years of measurements were analyzedand presented in Kalirai et al. 2011. We present the continuation of that analysis for allrelevant data taken up through August 2020. In particular, we analyze long-term changes tothe sensitivity of the IR channel, as accurately quantifying the sensitivity of the instrumentis crucial to the computation of the zeropoints. We also present some insights regarding thelower than expected precision of photometry with the IR channel. Furthermore, we discuss

Copyright© 2021 The Association of Universities for Research in Astronomy, Inc. All Rights Reserved.

https://www.stsci.edu/hst/instrumentation/wfc3/data-analysis/photometric-calibration/ir-photometric-calibration


the limitations to the photometric calibration due to the use of standard star images, andpossible future studies that can be made to achieve increased precision. Lastly, we brieflydiscuss the impact of updated flatfield calibrations on the standard star images and thederived photometry.

Observations

Due to their highly stable nature, and well understood physics, the HST standard stars areobserved by a large volume of both calibration and general observer proposals. We use allimaging data of these stars in this analysis, covering all 15 WFC3/IR filters, save for thecases in which the point spread function (PSF) falls off the boundary of the subarray ofactive pixels, or where the star is saturated. As some of the programs were not designed forphotometric calibration purposes, the number of observations for each target, and filter ishighly variable. The list of programs in which data were taken for each star is presented inTable 1.

Star Proposal IDs

GD-153 11451, 11552, 11926, 12334, 12699, 12702, 13089, 13092, 13575, 13579, 13711,

14021, 14384, 14386, 14544, 14883, 14992, 14994, 15113, 15582, 16030

GD-71 11926, 11936, 12333, 12334, 12357, 12699, 12702, 13711, 14024, 14384, 14883,

14992, 15113, 15582, 16030

GRW+70 5824 11557, 12333, 12698, 13088, 13575, 15582, 16030

P330E 11451, 11926, 12334, 12699, 13089, 13573, 13575, 14021, 14328, 14384,

14883, 14992, 16030

G191B2B 11926, 12334, 13094, 13576, 13711, 15113

Table 1: Proposal IDs for observations used in these analyses. For more information, see theMAST archive.

The majority of the datasets used in this analysis were observed as part of photometriccalibration programs, and typically feature large enough exposure times to exceed a signal-to-noise ratio (SNR) greater than 100. In more recent proposals (starting in 2017) theobservations were designed with persistence mitigation techniques in mind. Specifically,frequent dithering to place the star on a recently unused portion of the detector was used tomitigate persistence. Reducing the effects of persistence is critical to achieving high precisionphotometry (Bajaj 2019).

Analysis

The data were downloaded from the MAST archive in August 2020 to ensure the most recentreference files were used. However, a new flatfield calibration was developed contemporane-ously (Mack et. al, 2020, in prep) and the resultant new flat files were used in the calwf3

2


processing (see the “Effects of New Calibration” section). Images were then grouped bytarget and filter. Within those groups, images taken within the same visit were drizzledtogether, as images taken in the same visit typically have very precise relative astrometry.This served to reduce the number of discrepant artifacts in the images. Source finding wasperformed on these drizzled images using the python package photutils implementation ofthe DAOFIND algorithm (Stetson 1987). The DAOFIND full width at half max was setto the approximate width of the IR PSF of approximately 1.2 pixels. Though the FWHMvaries slightly with filter pivot wavelength a parametrization with respect to wavelength wasunnecessary for satisfactory results. Due to the highly undersampled nature of the IR PSF,many spurious sources would often be detected. In some cases, due to larger subarray usageand longer exposure times, other sources may also appear in the images, leading to extradetections.

To dispense of the superfluous detections, an initial pass of aperture photometry wasperformed on the images. The measured countrates were then compared to synthetic valuescomputed by PySynphot, using empirical models of the standard stars and total systemthroughput curves. The object that reported the closest countrate to the synthetic countratewas used to record an approximate position of the standard star in each image. This provedto be the most successful out of many source detection methods that were implemented, asthe synthetic flux values and photometric performance are consistent enough (within a fewpercent) to ensure an accurate selection.

The approximate positions from the drizzled images were then transformed back to theFLT image coordinate system via the all_pix2world() and all_world2pix() methodsof the astropy WCS package (The Astropy Collaboration et al. 2018). The positions werethen recentered in each image using 2D gaussian fitting to the central-most pixels of thePSF, ensuring that the small aperture used in the photometry is placed correctly. Typicalaperture photometry is then performed on the pixel-area-map-corrected FLT images, usingan aperture radius of 3 pixels (0.4”) and a background annulus ranging from 15 to 30pixels, using a sigma clipped median to calculate sky level. Unlike the analysis in Kalirai(2011), the pixel area map multiplication was necessary, as the placement of the stars onthe images spanned much of the total detector area. Thus, the pixel area map was sliced tocontain only the pixels used in each subarray image, and multiplied by the data array beforephotometry was performed. The aperture photometry was accomplished using photutilsand the wfc3 photometry package (Bradley et al. 2017). Since the FLT images for the IRchannel are already reduced to electrons per second, no further corrections were required. InKalirai et al. 2011, the aperture radius of three pixels was used, but also was described asnot being optimal for minimizing the dispersion of the measurements. However, in repeatingthis analysis with more data, we find that the three pixel aperture minimizes the standarddeviation for the GD-153 (the most observed star of the set) flux measurements in both theF110W and F160W filters.

3


Figure 1: Comparison of the normalized standard deviation for F110W and F160W as afunction of aperture radius. Since an aperture radius of three pixels minimized dispersionfor both short and long wavelengths, this radius was used for all filters in the analysis

Photometry

In general, the photometry of the IR detector remains stable over time. The (three sigma)clipped standard deviation of the flux measurements of a given target/filter were normalizedby the median flux measurement to compare the photometry across the various standards.This normalized standard deviation (in percent) is presented in Table 2. The updated ze-ropoints from these measurements are shown in appendix A, and lots showing photometryover time for all five standards in the 15 WFC3/IR filters are shown in appendix B.

4


Filter GD-153 GD-71 P330E GRW+70 5824 G191B2B

F098M 0.99 1.69 1.07 1.17 1.76

F105W 1.84 1.92 1.16 1.19 1.41

F110W 0.96 1.2 1.01 0.63 1.31

F125W 1.14 1.16 1.03 1.58 0.72

F126N 1.95 0.91 1.05 0.65 0.22*

F127M 1.19 1.03 0.98 0.83 1.36

F128N 1.77 0.94 1.26 0.59 0.1*

F130N 2.3 0.84 1.16 0.39 0.13

F132N 2.49 0.96 1.19 0.29 0.01*

F139M 1.69 0.93 1.0 0.99 0.99

F140W 0.94 1.22 0.88 0.72 0.89

F153M 1.59 0.75 0.71 0.6 0.72

F160W 0.85 0.96 0.92 0.72 0.71

F164N 2.13 1.13 1.14 0.47 0.18*

F167N 2.04 1.04 1.07 0.63 0.05*

Table 2: The normalized standard deviation of the photometric measurements of the stan-dard stars, in percent. The entries labeled with ”*” have less than 3 data points, and shouldnot be considered as meaningful.

5


The Search for Systematics

While the dispersion for many of the filters is fairly small (less than 1%), the signal-to-noise ratio of many of the observations is often substantially larger than 100, even includingnoise imparted from calibration (as reported in the error array of the FLT images). Figure 2shows the how the dispersion of photometry evolves with signal to noise ratio (flux divided byphotometric error) of the exposures. Notably, the actual standard deviations are consistentlyhigher than predicted for all SNR levels.

Figure 2: The normalized standard deviation (in percent) of photometric measurements ofGD-153 vs signal-to-noise ratio. The blue line represents the expected normalized standarddeviation (1/SNR).

6


In some cases, inclusion of images taken with different observation strategies imparts ahigher dispersion onto the photometry. Some of the observations of GD-153 F105W, forexample, were used in the WFC3/IR grism calibration and only include a small numberof reads per exposure, resulting in much more noisy data, due to the poorly understoodbehavior of the first read of WFC3/IR integrations (see Figure 3). Removing these lowsample exposures from the analysis increases precision for a small subset of the filters, thoughnot to the level predicted by the SNR. In the example of the GD-153/F105W images withless than 6 reads have a clipped standard deviation of 2%, while those with more reads havea much smaller dispersion of 0.7%. In addition, the difference between the means of the twopopulations differ by approximately 1.3%. However, this does not always yield a more preciseresult, and actually increases the dispersion for some other target/filter combinations.

Figure 3: : Flux of GD-153 taken in F105W. Points are color coded by number of samples(nsamp). The yellow points are NSAMP


and are therefore used in photometric calibration programs since 2017. .However, the WFC3/IR detector also exhibits longer term behavior, where even the

first observations in a visit (which should be unaffected by persistence) show photometricoffsets compared to previous visits (Bajaj 2019). In some cases, these offsets are presentacross a visit. The visit-to-visit variation is distinct from the Poisson error, as Poisson errorsmanifest randomly. This effect is also detected in spatial scan data, where Poisson noiseterms are effectively 0 (Som 2020, in prep). This instability between visits is not currentlywell understood.

Effects of New Calibration

An additional factor that differentiates this analysis and the resulting zeropoints from thoseof Kalirai et al. 2011 is the new calibration. The updated flat fields described in Mack et.al (2020, in prep.), while incorporating more data, do not substantially reduce the scatterof the measurements. This is partially due to the clustering of observations of the standardstars near the center of the detector (as WFC3/IR subarrays are centered on the array).The error in the flatfield in the center of the detector was already below the half percentlevel (Dahlen 2013), and thus the change in the new flats pixel to pixel variation is minimal.However, a change to the normalization of the flats causes a 0.1% drop in calibrated countrates, which is compensated for via a corresponding change in the zeropoint. Thus, whenthe zeropoint is applied, the measured flux is the same between calibration versions.

Though the photometry is consistent with the results of Kalirai et al. 2011, the zeropointsfor WFC3/IR change by a small amount due to recomputation of the synthetic models ofthe standard white dwarf stars. A small, wavelength-dependent update to the CALSPECmodels yields a slightly different zeropoint, as it represents a change of the physical fluxincident on the detector (Bohlin, Hubeny, and Rauch 2020). Though the change in thezeropoints would appear to show a loss of sensitivity (a brighter magnitude), it should beinterpreted as a more accurate, higher physical flux estimate of the sources measured, andnot a sensitivity loss. The total change of the zeropoints is presented in Figure 4.

Discussion and Supplemental Studies

The overall stability of the detector appears to remain similar to the results found in Kaliraiet al. 2011, with a typical dispersion of σ ≈ 1% and no significant trends consistent acrosstargets or fiilters. However, lack of precision and nonrepeatability of the photometric mea-surements ultimately limit the ability to detect the small sensitivity losses on the order ofthose seen in other HST instruments. Specifically, the visit-to-visit variation of the photome-try substantially reduces the precision of any time dependent measurement of the sensitivity.Thus, the measurements taken in the WFC3/IR sensitivity monitoring programs are unableto support the findings seen in other studies such as Kozhurina-Platais and Baggett 2020,which seem to detect sensitivity losses on the order of 1-3% over 10 years using observationsof the core of Omega Centauri.

A portion of the nonrepeatability may be attributed to varying observation configura-tions (e.g. different sample sequences, number of samples, and exposure time). A substantial

8


Figure 4: : Change in the STMAG zeropoints for WFC3/IR. The systematically largerchange in the redder wavelengths is primarily due to the new synthetic models.

detection or correction of systematic behavior as a function of these observation character-istics would likely require additional, extensive processing in the calibration pipelines. Thedependence of the photometry on these characteristics is currently being investigated.

Current calibration programs seek to measure sensitivity losses via spatial scanning, asthis observation strategy allows for extremely small Poisson noise terms. However, prelimi-nary analysis shows uncertainties much larger than the Poisson noise would predict withina visit, and from visit to visit (Som, 2020 in prep). This effect is not persistence related alsonot currently well understood, but appears consistent with the visit to visit variability of thestandard star measurements. However, these observations were designed to limit persistenceeffects, and have a much more unified observation strategy (compared to the standard starobservations), reducing parameter space in the analysis of systematics.

Observations of other, less crowded stellar clusters with well designed, consistent strate-gies between epochs may yield more precise measurements of sensitivity losses, and arebeing considered for future calibration cycles. We also seek to support measurements of thetime-dependent sensitivity via analyses of archival data, though heterogeneity in observationtargets and strategies may limit the efficacy of those studies.

Updated Calibration Reference Files

The changes to the calibration of WFC3/IR data are encapsulated in subset of the refer-ence files available in the Calibration Reference Data System (CRDS). The new flatfieldcalibrations are stored in new PFLTFILE (pixel-to-pixel flat file) and DFLTFILE (delta flatfile, storing information for the time dependent IR blobs) files. More information aboutthese files is available in Mack et. al (2020, in prep). The new zeropoints are stored in

9


a new Image Photometry Table (IMPHTTAB) as inverse sensitivities (PHOTFLAM). Notethat these zeropoints are computed for a photometric aperture of infinite radius. Thesefiles are necessary for data recalibration consistent with the new zeropoints. The new refer-ence file, 4af1533ai_imp.fits, is now available on the HST CRDS website. To maintainconsistency with the IMPHTTAB, the filter curves used in the construction of bandpasses forsynthetic photometry (PySynphot/STSynphot) were updated to reflect the new sensitivity.More information regarding the download of these files is available here. The models of thestandard stars used (Bohlin, Hubeny, and Rauch 2020) in this analysis were also updatedand placed in the CALSPEC data repository, and are available here. Note that these filesare not necessary for the recalibration of images, and are solely for synthetic photometry.

Acknowledgements

We thank A. Calamida, H. Khandrika, J. Mack, J. Medina, and D. Som of the WFC3Photometry Calibration group for their suggestions regarding analysis contained in andclarity of this report. We also thank J. Green for his review of this document.

References

Bajaj, V.. (June 2019). WFC3/IR Photometric Repeatability. WFC3 ISR 2019-07.Bohlin, Ralph C., Ivan Hubeny, and Thomas Rauch (July 2020). “New Grids of Pure-

hydrogen White Dwarf NLTE Model Atmospheres and the HST/STIS Flux Calibra-tion”. In: 160.1, 21, p. 21. doi: 10 . 3847 / 1538 - 3881 / ab94b4. arXiv: 2005 . 10945[astro-ph.SR].

Bradley, Larry et al. (Oct. 2017). astropy/photutils: v0.4. doi: 10.5281/zenodo.1039309.url: https://doi.org/10.5281/zenodo.1039309.

Dahlen, T. (Jan. 2013). WFC3/IR Spatial Sensitivity Test. WFC3 ISR 2013-01.Kalirai, J. et al. (Feb. 2011). The Photometric Performance of WFC3/IR: Temporal Stability

Through Year 1. WFC3 ISR 2011-08.Kozhurina-Platais, V. and S. Baggett (Apr. 2020). WFC3 IR sensitivity over time. WFC3

ISR 2020-05.Long, K., S. Baggett, and J. Mackenty (July 2013). Characterizing Persistence in the WFC3

IR Channel: Finite Trapping Times. WFC3 ISR 2013-06.Ryan, R. and S. Baggett (July 2015). The Internal Flat Fields for WFC3/IR. WFC3 ISR

2015-11.Stetson, Peter B. (Mar. 1987). “DAOPHOT: A Computer Program for Crowded-Field Stellar

Photometry”. In: 99, p. 191. doi: 10.1086/131977.The Astropy Collaboration et al. (Jan. 2018). “The Astropy Project: Building an inclusive,

open-science project and status of the v2.0 core package”. In: ArXiv e-prints. arXiv:1801.02634 [astro-ph.IM].

10

https://hst-crds.stsci.edu/https://www.stsci.edu/hst/instrumentation/reference-data-for-calibration-and-tools/synphot-throughput-tableshttps://www.stsci.edu/hst/instrumentation/reference-data-for-calibration-and-tools/astronomical-catalogs/calspechttps://doi.org/10.3847/1538-3881/ab94b4https://arxiv.org/abs/2005.10945https://arxiv.org/abs/2005.10945https://doi.org/10.5281/zenodo.1039309https://doi.org/10.5281/zenodo.1039309https://doi.org/10.1086/131977https://arxiv.org/abs/1801.02634


Appendices

A Zeropoints

Filter Pivot PhotBW ABMag Vegamag STmag mag error PHOTFLAMF098M 9864.72 500.85 25.6661 25.0984 26.9445 0.008 6.0653e-20F105W 10551.05 845.62 26.2637 25.6122 27.6882 0.0047 3.0507e-20F110W 11534.46 1428.48 26.8185 26.0528 28.4364 0.0046 1.5318e-20F125W 12486.06 866.28 26.2321 25.3248 28.0221 0.0078 2.2446e-20F126N 12584.89 339.31 22.8491 21.9218 24.6563 0.0079 4.9671e-19F127M 12740.29 249.56 24.6246 23.6572 26.4584 0.0122 9.4524e-20F128N 12831.84 357.44 22.9561 21.9129 24.8055 0.0078 4.348e-19F130N 13005.68 274.24 22.9813 21.999 24.8599 0.0081 4.1416e-19F132N 13187.71 319.08 22.9325 21.9289 24.8413 0.0071 4.2096e-19F139M 13837.62 278.02 24.4663 23.3815 26.4796 0.0043 9.3134e-20F140W 13922.91 1132.38 26.4502 25.3676 28.4768 0.0068 1.4759e-20F153M 15322.05 378.95 24.4469 23.187 26.6815 0.006 7.7161e-20F160W 15369.18 826.25 25.9362 24.6785 28.1774 0.0086 1.9429e-20F164N 16403.51 700.06 22.8921 21.5013 25.2748 0.0121 2.8257e-19F167N 16641.6 645.24 22.9366 21.5686 25.3505 0.0095 2.6215e-19

Table 3: The updated WFC3/IR zeropoints. The pivot and PhotBW columns are inAngstroms, while the PHOTFLAM column is in ergs/(cm2*Angstrom*s*e−)

11


B Photometry plots

In the following plots, five panels represent flux measurements for G191B2B, GD-153, GD-71, GRW+70 5824, and P330E, respectively. The measurements have been normalized bydividing the flux measurements by the median for that star/filter.

0.975

1.000

1.025F098M Stability

0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

12


0.975

1.000

1.025F105W Stability

0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

13


0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

0.975

1.000

1.025F126N Stability

0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

14


0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

15


0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

16


0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

17


0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

18


0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

0.975

1.000


0.975

1.000

1.025

0.975

1.000

1.025

Relativ

e Flux

0.975

1.000

1.025

55000 55500 56000 56500 57000 57500 58000 58500 59000Date [MJD]

0.975

1.000

1.025

19

AppendicesZeropointsPhotometry plots

Documents

Updated WFC3/IR Photometric Calibration...2020/12/02 · Calibration V. Bajaj, A. Calamida, J. Mack December 2, 2020 Abstract We present the continued analysis of photometric measurements